7.2.2.1 General design of cloud schemes within climate models

The potential complexity of the response of clouds to climate change was identified in the SAR as a major source of uncertainty for climate models. Although there has been clear progress in the physical content of the models, clouds remain a dominant source of uncertainty, because of the large variety of interactive processes which contribute to cloud formation or cloud-radiation interaction: dynamical forcing – large-scale or sub-grid scale, microphysical processes controlling the growth and phase of the various hydrometeors, complex geometry with possible overlapping of cloud layers. Most of these processes are sub-grid scale, and need to be parametrized in climate models.

As can be inferred from the description of the current climate models gathered by AMIP (AMIP, 1995; Gates et al., 1999) the cloud schemes presently in use in the different modelling centres vary greatly in terms of complexity, consistency and comprehensiveness. However, there is a definite tendency toward a more consistent treatment of the clouds in climate models. The more widespread use of a prognostic equation for cloud water serves as a unifying framework coupling together the different aspects of the cloud physics, as noted in the SAR. The evolution of the cloud schemes in the different climate models has continued since then.

The main model improvements can be summarised as follows:

(i) Inclusion of additional conservation equations representing different types of hydrometeors
A first generation of so-called prognostic cloud schemes (Le Treut and Li, 1991; Roeckner et al., 1991; Senior and Mitchell, 1993; Del Genio et al., 1996), has used a budget equation for cloud water, defined as the sum of all liquid and solid cloud water species that have negligible vertical fall velocities. The method allows for a temperature-dependent partitioning of the liquid and ice phases, and thereby enables a bulk formulation of the microphysical processes. By providing a time-scale for the residence of condensed water in the atmosphere, it provides an added physical consistency between the respective simulations of condensation, precipitation and cloudiness. The realisation that the transition between ice and liquid phase clouds was a key to some potentially important feedbacks has prompted the use of two or more explicit cloud and precipitation variables, thereby allowing for a more physically based distinction between cloud water and cloud ice (Fowler et al., 1996; Lohmann and Roeckner, 1996).

(ii) Representation of sub-grid scale processes
The conservation equations to determine the cloud water concentration are written at the scale explicitly resolved by the model, whereas a large part of the atmospheric dynamics generating clouds is sub-grid scale. This is still an inconsistency in many models, as clouds generated by large-scale or convective motions are very often treated in a completely separate manner, with obvious consequences on the treatment of anvils for example. Several approaches help to reconcile these contradictions. Most models using a prognostic approach of cloudiness use probability density functions to describe the distribution of water vapour within a grid box, and hence derive a consistent fractional cover (Smith, 1990; Rotstayn, 1997). An alternative approach, initially proposed by Tiedtke (1993), is to use a conservation equation for cloud air mass as a way of integrating the many small-scale processes which determine cloud cover (Randall, 1995). Representations of sub-grid scale cloud features also require assumptions about the vertical overlapping of cloud layers, which in turn affect the determination of cloud radiative forcing (Jacob and Klein, 1999; Morcrette and Jakob, 2000; Weare, 2000a).

(iii) Inclusion of microphysical processes
Incorporating a cloud budget equation into the models has opened the way for a more explicit representation of the complex microphysical processes by which cloud droplets (or crystals) form, grow and precipitate (Houze, 1993). This is necessary to maintain a full consistency between the simulated changes of cloud droplet (crystal) size distribution, cloud water content, and cloud cover, since the nature, shape, number and size distribution of the cloud particles influence cloud formation and lifetime, the onset of precipitation (Albrecht, 1989), as well as cloud inter-action with radiation, in both the solar and long-wave bands (Twomey, 1974). Some parametrizations of sub-grid scale condensation, such as convective schemes, are also complemented by a consistent treatment of the microphysical processes (Sud and Walker, 1999).

In warm clouds these microphysical processes include the collection of water molecules on a foreign substance (hetero-geneous nucleation on a cloud condensation nucleus), diffusion, collection of smaller drops when falling through a cloud (coalescence), break-up of drops when achieving a certain threshold size, and re-evaporation of drops when falling through a layer of unsaturated air. In cold clouds, ice particles may be nucleated from either the liquid or vapour phase, and spontaneous homogeneous freezing of supercooled liquid drops is also relevant at temperatures below approximately –40°C. At higher temperatures the formation of ice particles is dominated by heterogeneous nucleation of water vapour on ice condensation nuclei. Subsequent growth of ice particles is then due to diffusion of vapour toward the particle (deposition), collection of other ice particles (aggregation), and collection of supercooled drops which freeze on contact (riming). An increase in ice particles may occur by fragmentation. Falling ice particles may melt when they come into contact with air or liquid particles with temperatures above 0°C.

Heterogeneous nucleation of soluble particles and their subsequent incorporation into precipitation is also an important mechanism for their removal, and is the main reason for the indirect aerosol effect. The inclusion of microphysical processes in GCMs has produced an impact on the simulation of the mean climate (Hahmann and Dickinson, 1997).

Measurements of cloud drop size distribution indicate a significant difference in the total number of drops and drop effective radius in the continental and maritime atmosphere, and some studies indicate that inclusion of more realistic drop size distribution may have a significant impact on the simulation of the present climate (Hahmann and Dickinson, 1997).

IPCC Homepage